Is investor belt-tightening brought on by COVID-19 the bitter medicine needed to bring sanity to the artificial intelligence funding bubble, flush out fakes, and shift venture capitalists’ attention to the most deserving companies?

In 2019, private investments in AI totaled over $70 billion, with startups taking $37 billion of that. Despite these figures, AI in real life is playing catch-up with the AI of popular imagination. Startups themselves have done much to stoke great expectations in the minds of the public.

Everywhere we look, we see companies promising cutting-edge AI capable of driving cars, designing software, selling merchandise — overall doing our jobs for us. But many are failing to deliver the goods, and this should make them, and everyone involved in AI, nervous. 

Over the years, the reality behind the AI fantasy has encompassed Uber’s fatal autonomous car crash; ambitious efforts of well-funded companies such as IBM falling short of initial objectives; 70% of executives reporting little or no impact from AI investments; self-described AI companies employing human workers to do the real tasks behind the scenes; a widely publicized lawsuit against one startup for touting AI capabilities it didn’t have; 40% of “AI” companies in Europe having no AI in their products at all. 

Artificial AI

AI washing — the practice of stamping “AI” on the packaging of a product whose real AI capabilities are dubious — has been taking place for years. Another way in which companies may attempt to cash in on AI hype is known as “pseudo AI.” This occurs when companies have human workers do tasks that customers believe AI is doing, which has been uncovered at numerous companies. For example, behind its purportedly AI-powered virtual assistant, M, Facebook had human workers performing errands for users before discontinuing it due to the obvious difficulty and expense involved in scaling it. Originally created in 2015, M was shut down by January 2018.

Similarly, Clara Labs and X.AI were found to be relying on human workers for their scheduling services, according to a 2016 Bloomberg report. Practices such as these are not only dishonest, but also raise serious privacy questions for users who believe they are chatting with a non-human bot. Would Clara’s customers have used the service if they knew someone was viewing their Uber receipts complete with pick-up and drop-off addresses? 

It’s possible that some companies fib about AI-performing tasks that, in reality, humans do in order to attract funding from investors seeking to make a return on a heavily touted technology trend. This was apparently the case at Engineer.ai/Builder.ai — a startup claiming to design software applications with AI, while the bulk of work is done by developers in India and other locations. Its former chief business officer, Robert Holdheim, sued the company in 2019 for exaggerating its technological capabilities. The case drew significant media attention. Holdheim told The Verge the company reeled in investors with AI capabilities that it had barely begun attempting to develop.

Is it a defensible business practice to promote premature AI features in order to secure funding to bring them into reality? Interestingly, even some large, established companies with plenty of cash and talent to put toward AI projects have yet to hit their AI targets. For example, in 2013 IBM first launched Watson for Oncology with the stated goal “to eradicate cancer,” sadly missing the mark on advising doctors for treatments after years of investment.

Alphabet Inc.’s DeepMind — a project aimed at developing artificial general intelligence comparable to a thinking human — spent over $1 billion over the past three years and yet has trouble with passing high school math.

If these Goliaths still face so many obstacles, what chance does a seed-round small fry have? And how much more disappointment can investors and customers take before pushing back? 

Will the real AI please stand up?

Despite failures, many industry analysts believe AI is ultimately worth fighting for. Swift analysis of very large data sets is at the core of much modern technological innovation, and it isn’t possible at scale without some variant of AI or machine learning. Companies doing difficult research and development to develop authentic AI and advance worthwhile AI use cases deserve big investor dollars, according to Dr. Tommy Weir, founder and chief executive officer of Enaible Inc., a startup offering AI for improved leadership and productivity. 

In fact, as long as we’re talking about “real AI,” then investments in the space over the past several years aren’t excessive at all, according to Weir in an interview with theCUBE. “We need to be pouring millions into these companies,” he said. 

The problem of blurred lines between real AI and “window-dressing AI” appears to be widespread, confusing the public and investors alike. One would think companies attempting to sell something, and those investing their cash in it, would take the time to fully understand it, but this is not always the case in the frenzied AI space.

“I’m stunned by how many people know the buzzwords but don’t actually know what AI is,” Weir said. “We need to push the envelope and encourage new development and use cases.”

The industry also needs a deep appreciation at the algorithmic level of what kind of AI can actually achieve this, and then invest generously in it, according to Weir, who added that real AI is actually pretty simple to define. It can be determined by answering one question: Are the algorithms core to the technology, or are they added on so the company is able to say there is AI?

A warning sign to be on the lookout for, according to Weir, is companies with “AI” all over their branding that are simply reapplying old algorithms they did not originate to new problems.

How did we wind up in an AI bubble with investors sinking billions into startups that don’t meet these standards? “That’s a fantastic question, and one that we grapple with in real time as we are in the midst of raising another round,” Weir said.

Some investors just don’t spend the time to see what’s under the hood and should spend more time investigating the algorithms and looking for original research and development, according to Weir. Also, if a company’s AI has a chance to work in real life in a way investors and the general public can see, when it delivers real-world results it will stand out from the fakes and get the attention it deserves. 

As it happens, COVID-19 may be hastening both of these conditions — through economic austerity pinching investors and through an urgent call to action for AI companies to help fight the pandemic and its numerous adverse effects. 

Coronavirus test for AI companies

Investment in AI startups dipped to a near two-year low of $5.6 billion in Q4 2019 amid the COVID-19 outbreak, according to CB Insights. The rebound up to $8.4 billion in Q1 2020 was due to Waymo LLC’s $2.3 billion mega round for its self-driving car division. Milk and honey will not flow so freely for young companies seeking funding in the near future, according to Ron Schmelzer, principal analyst at AI-focused research and advisory firm Cognilytica Inc. 

“AI startups are being hammered. VC investment has slumped significantly,” Schmelzer told theCUBE. “The anecdotal information we’re hearing is that new rounds are going to be much tougher in Q2.” 

Economic uncertainty, layoffs, and mobility constraints are already impacting the AI ecosystem. It makes sense to expect that investors will be looking more closely at whether AI companies are worth investing in or not. Those who will fare the best are companies offering AI capabilities that can ease the burden of the pandemic in some way — for instance, remote work and productivity, EdTech, logistics and delivery, and healthcare companies. 

“While COVID-19 impacted AI deals in Q1, especially seed-stage and angel rounds, we do not expect AI funding and R&D to dry up in the long term,” Victor Adeleke, associate analyst at CB Insights, told theCUBE. “Rather, investors may pour more capital into companies and industries where the current pandemic has exposed a greater need for automation. The industries that attracted the most AI deals in Q1’20, which includes healthcare, finance, and media, will continue to evolve.” 

AI has shown potential in healthcare for years, and this is an opportunity for it to bear tangible results for the whole world to see. Companies are now bringing forth their best efforts in a race to solve a very real problem. For example, Massachusetts-based Scipher Medicine has partnered with Northeastern University’s Barabasi Labs, Harvard Medical School, and the Network Science Institute in an attempt to rush a coronavirus treatment to market.

Rather than developing a brand new drug — which might take two years in the Byzantine system of pharmaceutical development and approval — Scipher is fusing AI and a method of molecular analysis called Network Medicine to determine whether existing drugs could treat the virus. The consortium said it was able to complete six months of work within three weeks and has identified 81 potential drugs. 

In London, BenevolentAI took just two days combing through vast research with AI tools to uncover a rheumatoid-arthritis drug, called baricitinib, as a possible treatment. The drug, manufactured by Eli Lilly, has both anti-inflammatory and antiviral qualities and is now going through an accelerated clinical trial conducted by the National Institutes of Health. Results are expected within a few months. 

The pandemic has presented AI companies an opportunity to secure funding from the government and other entities besides traditional investors. Enterprise-AI software startup C3.ai, in tandem with Microsoft Corp., the University of California, Berkeley, and a number of other institutions, recently established C3.ai Digital Transformation Institute (known as C3.ai DTI), a consortium devoted to “digital transformation of business, government and society.”

In its first call for proposals, it asked for applications of AI and machine learning to fight the pandemic, including tools for prevention, testing, and societal resilience. It will grant research rewards from its $367 million fund to the best candidates. The due date for proposals was May 1. 

Some expect AI and automation for the workplace, industry and productivity to see a bump due to economic constraints. In fact, C3.ai Chief Executive Officer Thomas M. Siebel recently stated that the company has seen a significant increase in sales across all sectors, even in the midst of the global COVID-19 shutdown. The company combines AI with low-code development for faster building and deploying of AI apps for diverse industries like finance, energy, manufacturing, etc. 

Reshuffling and rebirth

COVID-19 isn’t death to AI investing, but it is occasioning a shift in attention — and funding — away from companies with little more to show the world than press releases.

“I think 2020 will be the year we see widespread death of AI startups due to lack of funding, especially the ones that have not yet proven a business model,” Schmelzer said.

Directing funding down narrower but richer avenues may prevent widespread disillusionment with AI, or even a new AI winter, and enable more fruitful research and development. Dr. Weir believes AI is unstoppable, and COVID-19 is ushering in conditions that will in fact catapult investment and development to new heights. These include the remote-work movement, tele-everything, and companies’ need to augment productivity. 

“[Real AI companies] are the companies that will be able to make sense of the vast data infrastructures that we built over the past few years,” Weir said. “The hype has preceded the impact. … This will be the decade of AI investment and impact.” 

Photo: Micheile Henderson/Unsplash